Irix joblimits failure (was: Re: New snapshot)

[David Kaelbling]

Ben Lindstrom wrote:
> 
> Am I going to see the fix for IRIX that you promised?  Or am I going to
> have to do a code revert before 3.2 release?

Sorry, I've been busy trying to ship our own release the last couple
weeks.  I should be able to work on this again after the 10th.  If the
freeze for 3.2 is coming up please pull my changes.  I can re-submit
them when they are more robust.

	Sorry,
	David

> On Thu, 7 Mar 2002, David Kaelbling wrote:
> 
> > IRIX has a compatibility mechanism that lets you test for optional symbols (like jlimit_start) at run-time.  I think these patches will let all all IRIX 6.5 systems build images that will test for job limit support dynamically:
> >
> > --- ./configure.ac    Wed Feb 27 01:12:35 2002
> > +++ ../openssh-3.1p1/./configure.ac   Thu Mar  7 15:50:21 2002
> > @@ -115,7 +115,7 @@
> >       AC_DEFINE(WITH_IRIX_ARRAY)
> >       AC_DEFINE(WITH_IRIX_PROJECT)
> >       AC_DEFINE(WITH_IRIX_AUDIT)
> > -     AC_CHECK_FUNC(jlimit_startjob, [AC_DEFINE(WITH_IRIX_JOBS)])
> > +     AC_DEFINE(WITH_IRIX_JOBS)
> >       AC_DEFINE(BROKEN_INET_NTOA)
> >       ;;
> >  *-*-linux*)
> >
> >
> > --- ./openbsd-compat/port-irix.c      Tue Feb 19 15:02:49 2002
> > +++ ../openssh-3.1p1/./openbsd-compat/port-irix.c     Thu Mar  7 15:35:21 2002
> > @@ -3,13 +3,20 @@
> >  #if defined(WITH_IRIX_PROJECT) || defined(WITH_IRIX_JOBS) || defined(WITH_IRIX_ARRAY)
> >
> >  #ifdef WITH_IRIX_PROJECT
> > -#include <proj.h>
> > +# include <proj.h>
> >  #endif /* WITH_IRIX_PROJECT */
> >  #ifdef WITH_IRIX_JOBS
> > -#include <sys/resource.h>
> > -#endif
> > +# include <sys/resource.h>
> > +# include <optional_sym.h>
> > +# if !defined(JLIMIT_CPU)
> > +/* Simulate job limit support so we can still test for it at runtime. */
> > +typedef __int64_t jid_t;
> > +extern jid_t jlimit_startjob(char *, uid_t, char *);
> > +#  pragma optional jlimit_startjob
> > +# endif
> > +#endif /* WITH_IRIX_JOBS */
> >  #ifdef WITH_IRIX_AUDIT
> > -#include <sat.h>
> > +# include <sat.h>
> >  #endif /* WITH_IRIX_AUDIT */
> >
> >  void
> > @@ -27,10 +34,16 @@
> >  #endif /* WITH_IRIX_JOBS */
> >
> >  #ifdef WITH_IRIX_JOBS
> > -        jid = jlimit_startjob(pw->pw_name, pw->pw_uid, "interactive");
> > -        if (jid == -1)
> > -                fatal("Failed to create job container: %.100s",
> > +     if (_MIPS_SYMBOL_PRESENT(jlimit_startjob)) {
> > +       jid = jlimit_startjob(pw->pw_name, pw->pw_uid, "interactive");
> > +       if (jid == -1) {
> > +         if (errno == ENOPKG)
> > +           jid = 0;
> > +         else
> > +           fatal("Failed to create job container: %.100s",
> >                      strerror(errno));
> > +       }
> > +     }
> >  #endif /* WITH_IRIX_JOBS */
> >  #ifdef WITH_IRIX_ARRAY
> >          /* initialize array session */
> >
> > (Hopefully that didn't get line-wrapped.  If it did I can try again.)
> > The JLIMIT_CPU test is to try and recognize when we are compiling on a system without the job limit headers.
> >
> >       Thanks,
> >       David
> >
> > --
> > David KAELBLING <drk at sgi.com>     Silicon Graphics Computer Systems
> > 1 Cabot Rd, suite 250; Hudson, MA 01749           781.839.2157, fax ...2357
> > _______________________________________________
> > openssh-unix-dev at mindrot.org mailing list
> > http://www.mindrot.org/mailman/listinfo/openssh-unix-dev
> >

-- 
David KAELBLING <drk at sgi.com>	    Silicon Graphics Computer Systems
1 Cabot Rd, suite 250; Hudson, MA 01749	    781.839.2157, fax ...2357